subagentic.ai RSS X YouTube GitHub
AI-generated illustration

News

n8n’s Nod is a local agent that stops outside the ask

n8n Labs’ Nod is an Apache-licensed local agent with its own browser. It asks before acting outside the request and does not replace the n8n workflow product.

n8nnodopen-sourcepersonal-agents

n8n’s official account used November 12, 2026, to pose “why another agent?” — naming OpenClaw, Hermes, Instinct, and Grok Bot — and called it a fair question.

The post from @n8n_io is about trust, not a launch clock. “We wanted something we’d feel comfortable handing our own accounts to.” The team said it likes the openness of OpenClaw and Hermes and the convenience of Instinct and Grok Bot, and that Nod is an attempt to bring those together.

The product facts are in n8n Labs’ research-preview write-up and on the Nod site. Labs is a new team inside n8n that experiments outside the core product. Nod is a standalone app, separate from n8n and from the agents you can build inside n8n. You do not need an n8n account, and there is no server or cloud service to set up first. n8n says the preview does not change the direction of the workflow product. Work on n8n and the n8n agent continues as planned.

Shared processes stay on a shared n8n instance. Nod is the single-player side: one person, that person’s accounts, and jobs that are theirs alone, for a business or a household.

It runs on the user’s computer, open source under Apache 2.0, with the model provider the user chooses. The product page lists Mac on macOS 14 or later, Windows 11 in preview, and a Linux x64 Flatpak preview, or your own server for one person. The app is free to download. You pay the model provider — an API key, a ChatGPT Plus or Pro plan, or a key from a provider such as OpenAI, Anthropic, Google, or OpenRouter — or you connect an open-source model on your own machine.

Nod has its own computer, a locally running sandbox the Labs post describes as a private virtual machine, and its own browser. It works in apps you connect, including Gmail, Calendar, Drive, and Slack, and on websites with no integration. The write-up’s examples are ordinary errands: cancel a subscription and ask for a refund, watch a price, book a hotel only after you say go, download invoices from supplier portals, and copy numbers out of a dashboard that cannot export them. You keep one conversation, in the app or on Telegram or WhatsApp, while several jobs run in parallel. Repeating work, such as a morning brief or a price check, can run on a schedule. Nod also reads connected apps and offers jobs it could take.

The Labs write-up is where the permission check is described. Before Nod sends, pays, posts, or deletes anything you did not ask for, a separate check confirms you asked for it. A web page or an email cannot give it orders. That check is itself a model and can make mistakes, which is why Nod asks when your words do not cover an action. Saved logins and cards stay encrypted on the machine and are entered through private forms the model cannot read. Sign-ins, card use, in-app actions, and task cost are logged.

Staying on your machine is not the same as never leaving it. Database, files, memory, and saved logins stay on the computer or server running Nod. The model service receives what each task needs. App connections use n8n and Composio by default. Beyond update checks, n8n receives usage data to improve Nod, and that can be turned off in Settings.

Early testers found some tasks cost more than they expected. n8n says a recent release brought that down substantially, without stating a typical price. Published guardrails: every task shows its cost and a finished task ends with a receipt; you can set a monthly spending limit; Nod checks in once a task passes $5, and you can change that amount; each schedule shows an estimated monthly cost.

The preview is also explicit about gaps. On a laptop, Nod stops when the laptop sleeps. Each installation has one owner. It cannot make phone calls. A hosted version, phone calls, and a Nod shared by more than one person are on the roadmap.

Connecting an n8n instance is optional, so Nod can work with workflows you already have. The product page states the boundary in the other direction too: Nod is its own app, made by the team behind n8n, and it is not the n8n AI Assistant.

Install the preview for your platform, connect a model you already use, and hand it one job you would otherwise do by hand. Then tell n8n how it went, on the community forum post linked from the Labs write-up or with Send feedback in the app.

Sources